Author: Jason Williamson
Date: 01:24:33 04/13/01
Go up one level in this thread
On April 12, 2001 at 14:14:14, Paul Doire wrote: >On April 12, 2001 at 13:41:41, John Merlino wrote: > >>On April 12, 2001 at 11:58:05, Paul Doire wrote: >> >>>On April 12, 2001 at 11:23:00, John Merlino wrote: >>> >>>>On April 12, 2001 at 07:07:47, Daniel Chancey wrote: >>>> >>>>>Could you explain the problem of hash clearing again? >>>>> >>>>>Castle2000 >>>> >>>>First, you should know that the problem will go away with the next (and final) >>>>patch. >>>> >>>>Second, it's pretty simple. The hash table is cleared out at the beginning of >>>>every search. On my PIII-600, it took about 1/2 second (by my VERY ROUGH >>>>estimations) to clear a 32MB hash table. Therefore, in a G/5 that goes 80 moves >>>>(for example) a Chessmaster personality with a 32MB hash table will lose 40 >>>>seconds on its clock for nothing, since 31MB of the hash table is completely >>>>wasted. >>>> >>>>jm >>> >>>Hi John, >>>Interestingly enough,CM Default doesn't do as well against GT-2 >>>on same pc matches. However CM KKND 32H does defeat GT-2 and CM Default at least >>>50-60% of the time (entertainment games only... OK cruel posters). >>>Do you think that somehow the custom personality CM KKND 32H clears it's hash >>>tables better than Default? It doesn't make sense but the results seem to make >>>it appear that way.Why else would CM KKND have better results? >>> >>>Thanks for your time, >>>Paul >> >>Unknown, since I do not have enough information to answer the question. Are we >>still talking about G/5 games? How many games have been played and what's the >>actual tally (W/L/D)? >> >>Didn't your other post say that the default soundly beat KKND? I'm confused.... >> >>jm > >Hi John, Thanks for replying, > >My other post entitled "CM8K personality tourney results-120 games" G/5 >The results were for the top two as follows : >CMKKND 32H +10=8-6 14 >CM Default +7=12-6 13 > >although overall score was within one, I thought CM KKND 32H win ratio was >more impressive than CM Defaults, and it does much better against GT-2. Hmm, they didn't even play the same number of games....Anyway, a difference of one point with 24 games played isn't goin to give you a statisticaly relavent result. You will need several hundred games to get even close to that. And I suspect you will find little proof that the other personalities are substantially better then the deafult one. (Which is 2nd on my tournament rating list with some 150 games played, following Deep Fritz and ahead of Junior 6, Nimzo 8, and others.)
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.